Will AI Stock Trading Really Come True?
Reporting from Chinese Private Equity Communities by Interface News Author: Liu Lilong on 2023-08-29
Since the beginning of this year, with the rapid iteration of artificial intelligence products such as ChatGPT, AI stock trading has once again become a hot topic that attracts attention.
Especially when the well-known private equity giant Magic Square Quantitative announced the exploration of general artificial intelligence (AGI), and when Beijing Private Equity Zhizhishan Investment announced that it would launch AI-only products, there was a big discussion in the investment communities. Investor Li Jiancheng (pseudonym) feels unprecedentedly confused, saying: "Isn't it enough to be harvested by quantification? Will it be harvested by AI again in the future?"
Full of gimmicks and imagination, will AI trading really come true?
Quantitative Private Equity Giants Start the AI Battle
This year, Huanfang Quantitative, a multibillion quantitative private equity, issued a document stating that it would concentrate its resources and strength on artificial intelligence that serves the common interests of all mankind, and establish a new independent research organization to explore the nature of AGI.
When AI stock trading spreads quickly, and many netizens call out "Artificial intelligence is harvesting wool?" and "Does this mean fund managers are going to lose their jobs?"
However, Huanfang Quantitative CEO Lu Zhengzhe soon stood up and responded, "Exploring AGI is not for stock trading. What we do is actually a completely independent large-scale model research that has nothing to do with finance. Long-term social value is our focus. We have established a new team independently of investment, which is equivalent to starting a business for the second time. It is just that we were not in the Internet technology industry before and we are an outsider. We ourselves want to do things of greater value and transcend the investment industry, but it has been misinterpreted as AI stock trading.”
Although it denies that "the purpose of exploring AGI is to speculate in stocks," Magic Square Quantitative has been laying out AI for a long time.
Public information shows that as early as October 2016, Magic Square Quantitative’s first stock position generated by a deep learning algorithm model was launched for real trading. By the end of 2017, almost all of the company’s quantitative strategies had been calculated using AI models.
In December 2019, Huanquan AI was officially established, dedicated to research on algorithms and basic applications of AI. In March 2020, Huanquan invested nearly 200 million yuan in quantified total investment to build the "Yinghuo One" equipped with 1,100 high-end graphics cards. Used to provide computing power support for Magic Square’s AI research . In January 2021, "Yinghuo 2" computing center, which the company invested 1 billion in construction, was officially delivered. In 2022, "Yinghuo 2" achieved software and hardware architecture innovations of multi-800-port switch interconnection and core expansion subtrees, breaking through the first phase due to physical limitations, the computing power is doubled.
In fact, it is not just Magic Square Quantification. A review of Jiemian News found that many multibillion quantitative private equity giants are actively deploying in the AI field.
Nian Kong Technology started the actual operation of machine learning strategies in May 2018, and equipped servers with high-performance GPU clusters to fully guarantee the team's computing power. So far, the company's deep learning-based machine learning algorithms have fully replaced traditional statistical arbitrage strategies and are used in all stock neutral strategy products.
Jiukun Investment CEO Wang Chen has publicly stated that the company’s investment research follows an “AI native” approach. “We first define this investment strategy as a set of AI architecture, using AI as the underlying function to drive our entire research ideas, research methods, and the upgrade of research logic. Our current investment research is 100% driven by AI, and of course it is deeply superimposed on some of our experience in researching quantitative investment rules."
Mingshi Private Equity also directly regards AI as a link that goes hand in hand with "factors, optimization, risk control, and trading" and jointly serves as one of the five major links in the company's strategic research and development.
Also a multibillion private equity, founder & investment director Xu Shunan of Inno Assets , said in an interview with Jiemian reporters that most of the multibillion private equity strategies now include artificial intelligence models. Inno Asset began to use artificial intelligence models in real trading in 2018, and now it has been fully applied in various strategic fields.
In Xu Shunan’s view, artificial intelligence is very suitable for the field of quantitative investment. Quantitative investment is mainly based on statistics, and artificial intelligence models are a very advantageous statistical method, especially suitable for big data and non-linear fields. Therefore, appropriate artificial intelligence models have the basis for effective application in quantitative investments. Investment performance over the past few years has proven that artificial intelligence models can achieve excellent performance in quantitative investments.
Among major quantitative companies is particularly fierce in the “fight for talents.”
Jiemian News investigated into the case and found that on recruitment websites, many quantitative private equity giants such as Huanfang Quantitative, Jiukun Investment, Lingjun Investment, and Jiaqi Investment have all offered million-level annual salaries for AI-related jobs.
Faced with the aggressive entry of major quantitative companies into AI, many investors have the same concerns as Li Jiancheng. They generally believe that AI has already defeated human beings in the world Go champion and is gradually replacing it in fields such as medical care, law, and manufacturing, therefore it is only a matter of time before they defeat humans in stock trading.
However, in the view of Zhang Qiang, a private equity research director in Guangzhou, although there is no doubt about the advantages of artificial intelligence and major quantitative companies are also continuing to explore AI, there are two major obstacles that are difficult to overcome from research to application of AI in the investment field.
On the one hand, there are many factors that affect market performance, and not all factors can be digitized. This means that these factors that cannot be digitized can never be learned by machines. AI investment is destined to exist with part of the blind spot.
On the other hand, AI can indeed grasp the factors affecting the market faster and more comprehensively than humans. However, in fact, the volume of A-shares is large enough, and investors do not need to fully grasp all the information. As long as they can grasp the main contradictions, they can capture opportunities to profit from them. In the process of grasping the main contradiction, people are often able to respond in a more timely manner. Because AI needs to go through a certain amount of learning and accumulation first, its response will be relatively lagging, especially when faced with changes in market trends.
Small Private Equity Companies Not Powerful Enough To "Take The Pit"
Due to the need for a large amount of computing power and manpower to support, compared with the large quantitative private equity firms with deep financial resources, small private equity firms are obviously at a clear disadvantage in this battle for AI.
Even Li Jiancheng, who is very panicked about AI stock trading, actually doesn't take small private equity investors seriously. In his words, "This kind of thing is done with money, and it can't be done without some financial resources."
However, many small private equity firms are still actively embracing AI. This "ambition" is first reflected in product naming.
Data shows that the number of products with the word "smart" in their names among quantitative private equity funds has increased from 38 to 78 from 2020 to 2023.
Figure: Number and distribution of artificial intelligence private equity funds
Source: Compiled by Private Equity Ranking Network and Jiemian News
Among them, most private equity funds with intelligently named products have management scales ranging from 0 to 500 million yuan. The latest data shows that the number of private equity funds with a scale of 0 to 500 million accounts for more than 75%.
Figure: Proportion of scale of artificial intelligence funds ranging from 0 to 500 million yuan
Source: Compiled by Private Equity Ranking Network and Jiemian News
This year , Beijing-based Zhiyu Zhishan Investment was the first in the industry to announce that it would launch AI-managed products.
Zhiyu Zhishan Investment stated that it plans to arrange for the company's four researchers and an artificial intelligence-based robot (tentatively named Cybertron) to independently manage five different private equity funds. This product that will be independently managed by AI is called "Zhiyu Zhishan No. 1". While being managed by Cybertron (AI), it will be supervised by He Li, the general manager of Zhizhishan Investment.
As soon as the news came out, it immediately caused an uproar in the investment circle, including many voices of doubt from many aspects.
The next day, Zhizhishan Investment was "slapped in the face". The company issued another announcement on June 2, stating that after discussion and reflection by the team, it believed that the true meaning and value of Cybertron (AI) within the existing value investment system is limited to the reshaping and integration of the best value investment system, and is not mainly used for trading, because from the first principle, game-based trading is not the main source of profit for value investment, let alone the team's advantage.
At this point, the farce of AI-managed products has ended hastily, but related topics are still arousing heated discussions. The focus of the issue is nothing more than: Are AI-managed products really reliable ?
Investor Bai Xiaojie is skeptical about this. In recent years, he has experienced many "intelligent stock selection" functions, such as Oriental Fortune's intelligent stock diagnosis, Flush's Aiwencai, etc., but the overall winning rate is extremely unstable, " It’s hard to imagine letting AI manage products on its own, and the results will most likely be catastrophic.”
Xu Shunan believes that it is inaccurate to use artificial intelligence to independently manage fund products.
“If we use the artificial intelligence model developed by the research team to manage funds, then many institutions have already done so, because the artificial intelligence model is a special statistical model and is not essentially different from the traditional statistical model. "
"If it refers to a robot replacing investment managers, then this is a misunderstanding of the artificial intelligence model. Artificial intelligence models cannot replace people, because all artificial intelligence models are developed by people, and all improvements are also made by people. The most important factor in quantitative investing is always people, not machines.”
Nian Kong Technology pointed out that AI sole-managed products violate the contract of private equity funds. Generally speaking, the first person responsible for fund products is the fund manager. If it is replaced by a robot, there will be no way to bear the responsibility. Wang Xiao even explained humorously, "Because someone has to take the blame. You can't let a machine take the blame, right?"
What Is the Future of AI Stock Trading?
It may be difficult for AI to manage products alone, but it is actually playing an increasingly important role in the field of private equity investment.
According to Hu Ping, deputy director of quantification at the private equity Xuanyuan Investment , the application of artificial intelligence in A-share investment is very diverse. In traditional subjective investment methods, artificial intelligence provides many convenient tools to quickly help investment managers and researchers obtain and process information; in quantitative investment methods, the underlying technology of artificial intelligence can help investment researchers analyze data from more dimensions, market modeling and trading signals generation. Regardless of the investment methodology, artificial intelligence technology can improve the efficiency of investment research and improve the accuracy of market understanding.
According to Wang Xiao, the entire quantification process is divided into four or five parts. For example, after mining of the characteristics of a series of factors, it needs to be fitted into a unique predictive value. Currently, Niankong Technology mainly uses artificial intelligence in this step.
Mingshi Private Equity is more biased towards behavioral finance. The basis of application is mainly factor mining (interpretability) based on financial logic. The research methods are more academic logic. AI artificial intelligence mainly makes full use of machines to learn powerful nonlinear processing capabilities to integrate factors into signals.
Zhou Yitian, founding partner and co- CEO of Blackwing Assets, a multi-billion private equity firm, said that the company’s investment research system is intelligent, integrated, and centralized, with data team, factor team, modeling team, algorithm team, and trading team, each of which is a clear division of labor and efficient collaboration. At present, this investment research system is relatively mature. Artificial intelligence technology is embedded in the entire strategy research and development process, including data analysis, factor mining, income forecasting, portfolio optimization and other links.
However, artificial intelligence also has obvious shortcomings, which Zhang Qiang summarized as the following three points:
1. Data quality: Artificial intelligence has high requirements for data quality in stock trading. If the data quality is inaccurate, the prediction results will be unreliable.
2. Uncertainty: The stock market is an environment full of uncertainties, including policy changes, international relations, natural disasters, etc. These factors pose challenges to the predictive capabilities of artificial intelligence.
3. Overfitting problem: Just because an artificial intelligence model has performed well on past data does not mean that it will continue to perform well in the future. If the model overfits historical data, it may lead to misjudgment of the actual situation.
Jiemian News compared the return performance of quantitative products and artificial intelligence products since 2020. Over the past 43 months, the average return rate of quantitative products has been 44.45%, while the average return rate of artificial intelligence products has been only -2.14%, a sharp decline. Losing the income performance of quantitative products in the same period also shows to a certain extent that artificial intelligence does not contribute much to the income of private equity products at this stage.
Figure: Revenue from artificial intelligence products and quantitative product indices
Source: Private Equity Pai Pai Network, Chaoyang Sustainability, compiled by Jiemian News
Comparing the trends of the Shanghai Composite Index and the number of artificial intelligence products since 2020, it can be found that no matter how the market performs, the number of artificial intelligence funds that announce their net values is not large. Generally speaking, many private equity investors choose not to update their net worth because their performance is not optimistic.
With regards to applications of ChatGPT, which has become popular this year, to conduct quantitative research, Wang Xiao, chairman of Niankong Technology, told Jiemian News that the biggest problem at present is that OpenAI will limit traffic usage. If the traffic limit is exceeded every month, it will no longer be available. However, the current limit is far from meeting the needs of quantitative private placement.
Regarding whether private equity investment will be replaced by AI in the future, Jiemian News found during the interview that many people, like Li Jiancheng, believe that this is only a matter of time.
Others hold a negative view. Zhang Qiang and Bai Xiaojie firmly believe that it is impossible. They always firmly believe that "people can never be defeated by machines they create ."
The answer given by Hu Ping is also negative. He believes that some jobs will be replaced, especially repetitive labor with fixed patterns, such as the acquisition and cleaning of various data and information, algorithmic trading execution, etc.; but for those who require more creativity, artificial intelligence cannot yet replace humans in the work of abstract thinking. In essence, the current advantage of artificial intelligence lies in induction, while the advantage of humans lies in deduction.
As for where artificial intelligence will go in the future in the investment field?
Hu Ping believes that professional investors will continue to introduce the successful methodologies of artificial intelligence in various fields into their investments. Market micro-pricing is becoming more and more efficient, and the market will enter a weakly efficient market. Once an efficient market is formed, the role of current artificial intelligence will be significantly reduced.
Xu Shunan further stated that the artificial intelligence model is just a special statistical method, and its future development cannot ultimately escape the scope of statistical models. It is just that model research is getting deeper, wider, and more sophisticated.
(Li Jiancheng, Bai Xiaojie and Zhang Qiang are all pseudonyms in the article)